Optimisations in the sequencer inbox

[yahgwai]

• Don't use proxy ~44k
• Immutable max time variation, as well as rollup and bridge address ~10k
• postUpgradeInit required, see Move event and delayed inbox messages read to the bridge #87

Nitro changes required:

• SequencerBatchDelivered now emitted from the bridge, also the event args have changed to include the sequencer inbox address.
• The sequencer inbox is now immutable, so we will now deploy a new one when changes are required in the sequencer inbox
• The variant of addSequencerL2BatchFromOrigin without the prevMessageCount, newMessageCount and parameters have been removed, however it will continue to exist on the old sequencer inbox which wont be updated to have it removed. If this function abi was used for decoding historical logs we should instead do this with standard abi decoding.

[gzeoneth]

worth noting the uint64 cast below is not safe, this is an existing issue and will only be a problem with misconfiguration; tho we might also consider to have some sanity check here to make sure each value is within some bound

or we can just perform some check there, we already check for underflow but not overflow

[yahgwai]

I think it's fine to have maxtimevariation as uint64s right? I've updated to that now

[shotaronowhere]

The modifier refundsGas(IGasRefunder gasRefunder) in the GasRefundEnabled abstract contract is unchanged, however it still calculates the gas refund assuming the calling contract is a proxy (copying calldata to memory). Since the sequencerInbox is no longer a proxy, the refundsGas modifier miscalculates. Underpayment is a problem, but overpayment isn't a big issue (this case), however it's an unnecessary calculation (wasted gas).

I would either make the GasRefundEnabled contract delegate call aware, or avoid abstracting the gas refund logic as general purpose contract all together.

I presume a refactor of the gasRefunder is planned given the eip 4844 support #96 , just highlighting the need.

[DZGoldman]

should we include setSequencerInbox in here too?

[yahgwai]

Yeh, I thinkn that makes sense, will update.

[yahgwai]

The modifier refundsGas(IGasRefunder gasRefunder) in the GasRefundEnabled abstract contract is unchanged, however it still calculates the gas refund assuming the calling contract is a proxy (copying calldata to memory). Since the sequencerInbox is no longer a proxy, the refundsGas modifier miscalculates. Underpayment is a problem, but overpayment isn't a big issue (this case), however it's an unnecessary calculation (wasted gas).

I would either make the GasRefundEnabled contract delegate call aware, or avoid abstracting the gas refund logic as general purpose contract all together.

I presume a refactor of the gasRefunder is planned given the eip 4844 support #96 , just highlighting the need.

Great catch, updated the gasrefundenabled contract to be delegate call aware. 4844 gas refund changes will happen in a separate pr.

[gzeoneth]

#110